Mixtures of Gaussian Random Fields and Its Application in Analyzing fMRI Data

• Activity Number: 471 - Contemporary Statistical Methods for Imaging Data Analysis
• Type: Contributed
• Date/Time: Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Mozhdeh Forghani* and Khalil Shafie
• Companies: University of Northern Colorado and University of Northern Colorado
• Keywords: Random Field Theory; Gaussian Random Field; Euler Characteristic; Mixture of Gaussian Random Fields

Abstract

The problem of searching for the activation in a brain region through the images obtained from fMRI has been the subject of many studies. One approach to test the activation is based on the Random Field Theory. In this approach, the whole brain is mostly modeled as a Gaussian random field. The idea of using random field theory in testing fMRI is to obtain p-value through the use of global maximum of the sample field and the notion of expected Euler characteristic. However, not always assuming the whole brain as a Gaussian random field is appropriate. In this work, we model the whole brain as a mixture of two Gaussian random fields. Here, we have also used the expected Euler characteristic to obtain p-value and test the activation. The method is applied for a real data set as well as simulated data.